---
title: Paddle
slug: /bundles-paddle
---

import Icon from "@site/src/components/icon";
import PartialParams from '@site/docs/_partial-hidden-params.mdx';

<Icon name="Blocks" aria-hidden="true" /> [**Bundles**](/components-bundle-components) contain custom components that support specific third-party integrations with Langflow.

This page describes the components that are available in the **Paddle** bundle.

The **PaddleOCR** component calls the [PaddleOCR AI Studio](https://github.com/PaddlePaddle/PaddleOCR) async Job HTTP API to process images and PDFs. It supports two task types:

* **Document parsing**: Layout-aware extraction into Markdown (`PP-StructureV3`, `PaddleOCR-VL-1.6`). Use this for PDFs, scanned documents, and content where reading order and structure matter.
* **OCR**: Plain text recognition in scan order (`PP-OCRv5`, `PP-OCRv6`). Use this for images with simple text content.

The component talks to the API over HTTP and does not require the `paddleocr` Python SDK.

## Install the Paddle bundle

The **Paddle** bundle is included in the `lfx-paddle` Extension bundle, which is installed automatically as part of `uv pip install langflow`.

If you need to install it separately, run:

```bash
uv pip install lfx-paddle
uv run langflow run
```

To verify the bundle is loaded in your environment:

```bash
lfx extension list
```

## Prerequisites

* A [Baidu AI Studio](https://aistudio.baidu.com/account/accessToken) access token.

## PaddleOCR

The **PaddleOCR** component uploads one or more image or PDF files to the PaddleOCR AI Studio Job API, polls until each job completes, and returns the extracted text.

Supported file types: `png`, `jpg`, `jpeg`, `bmp`, `tiff`, `webp`, and `pdf`.

It outputs a [`Table`](/data-types#table) with one row per processed file. Each row includes fields such as `text`, `file_path`, `task_type`, `output_format`, `model`, `job_id`, and `pages`.

* For **document parsing**, `text` is Markdown and `output_format` is `markdown`.
* For **OCR**, `text` is plain text and `output_format` is `plain_text`.

### Use the PaddleOCR component in a flow

1. Get an access token from [Baidu AI Studio](https://aistudio.baidu.com/account/accessToken).
2. In Langflow, add the **PaddleOCR** component from the **Paddle** bundle.
3. Enter your token in the **AI Studio Access Token** field, or add it as a [global variable](/configuration-global-variables).
4. Upload one or more supported files, or connect a server file path.
5. Select a **Task Type**:

    * **document_parsing** for layout-aware Markdown using models `PP-StructureV3` or `PaddleOCR-VL-1.6`.
    * **ocr** for plain text recognition using models `PP-OCRv6` or `PP-OCRv5`.

6. Select a **Model**. The available models change based on the selected task type.
7. Connect a **Chat Output** or other downstream component to the **Files** output to inspect the extracted text.
8. Click <Icon name="Play" aria-hidden="true" /> **Playground** to run the flow.

### PaddleOCR parameters

<PartialParams />

| Name | Type | Description |
|------|------|-------------|
| **Files** (`path`) | File | Input parameter. One or more image or PDF files to process. Supported types: `png`, `jpg`, `jpeg`, `bmp`, `tiff`, `webp`, `pdf`. |
| **AI Studio Access Token** (`access_token`) | SecretString | Input parameter. Baidu AI Studio access token used to authenticate Job API requests. Required. |
| **Task Type** (`task_type`) | Dropdown | Input parameter. `document_parsing` for layout-aware Markdown, or `ocr` for plain text recognition. Default: `document_parsing`. |
| **Model** (`model`) | Dropdown | Input parameter. PaddleOCR model for the selected task type. For document parsing: `PP-StructureV3` (default), `PaddleOCR-VL-1.6`. For OCR: `PP-OCRv6` (default), `PP-OCRv5`. |
| **Base URL** (`base_url`) | String | Input parameter. Optional PaddleOCR service root URL. Leave empty to use the official default (`https://paddleocr.aistudio-app.com`). Advanced. |
| **Timeout (s)** (`poll_timeout`) | Integer | Input parameter. Maximum time in seconds to wait for each PaddleOCR job to complete. Default: `600`. Advanced. |
| **Document Orientation Classification** (`use_doc_orientation_classify`) | Boolean | Input parameter. Enable document orientation classification. Default: `false`. Advanced. |
| **Document Unwarping** (`use_doc_unwarping`) | Boolean | Input parameter. Enable document unwarping. Default: `false`. Advanced. |
| **Text Line Orientation** (`use_textline_orientation`) | Boolean | Input parameter. OCR option. Enable text line orientation detection. Default: `false`. Advanced. |
| **Text Detection Threshold** (`text_det_thresh`) | Float | Input parameter. OCR option. Text detection threshold. Advanced. |
| **Text Detection Box Threshold** (`text_det_box_thresh`) | Float | Input parameter. OCR option. Text detection box threshold. Advanced. |
| **Text Detection Unclip Ratio** (`text_det_unclip_ratio`) | Float | Input parameter. OCR option. Text detection unclip ratio. Advanced. |
| **Text Recognition Score Threshold** (`text_rec_score_thresh`) | Float | Input parameter. OCR option. Text recognition score threshold. Advanced. |
| **Table Recognition** (`use_table_recognition`) | Boolean | Input parameter. Document parsing option. Enable table recognition. Default: `true`. Advanced. |
| **Formula Recognition** (`use_formula_recognition`) | Boolean | Input parameter. Document parsing option. Enable formula recognition. Default: `false`. Advanced. |
| **Chart Recognition** (`use_chart_recognition`) | Boolean | Input parameter. Document parsing option. Enable chart recognition. Default: `false`. Advanced. |
| **Seal Recognition** (`use_seal_recognition`) | Boolean | Input parameter. Document parsing option. Enable seal recognition. Default: `false`. Advanced. |
| **Prettify Markdown** (`prettify_markdown`) | Boolean | Input parameter. Document parsing option. Return prettier Markdown when supported. Default: `true`. Advanced. |
| **Temperature** (`temperature`) | Float | Input parameter. PaddleOCR-VL option. Sampling temperature. Advanced. |
| **Top P** (`top_p`) | Float | Input parameter. PaddleOCR-VL option. Nucleus sampling `top_p`. Advanced. |
| **Visualize** (`visualize`) | Boolean | Input parameter. Document parsing option. Generate visualization outputs when supported. Default: `false`. Advanced. |
| **Server File Path** (`file_path`) | Data, JSON, or Message | Input parameter. Alternative to **Files**. A Data object with a server file path, or a Message with a path. Advanced. |
| **Separator** (`separator`) | String | Input parameter. Separator used between multiple outputs in Message format. Default: `\n\n`. Advanced. |
| **Silent Errors** (`silent_errors`) | Boolean | Input parameter. If `true`, errors do not raise an exception. Advanced. |
| **Delete Server File After Processing** (`delete_server_file_after_processing`) | Boolean | Input parameter. If `true`, deletes the server file path after processing. Default: `true`. Advanced. |
| **Ignore Unsupported Extensions** (`ignore_unsupported_extensions`) | Boolean | Input parameter. If `true`, files with unsupported extensions are skipped. Default: `true`. Advanced. |
| **Ignore Unspecified Files** (`ignore_unspecified_files`) | Boolean | Input parameter. If `true`, Data objects without a server file path are ignored. Default: `false`. Advanced. |
| **Files** (`dataframe`) | Table | Output parameter. One row per processed file, including extracted `text` and job metadata. |

## See also

* [PaddleOCR documentation](https://paddlepaddle.github.io/PaddleOCR/latest/en/version3.x/paddleocr_and_ppstructure.html)
* [Baidu AI Studio access token](https://aistudio.baidu.com/account/accessToken)
* [Langflow Extensions overview](/extensions-overview)